一种基于RBF神经网络预测模型的DMC预测控制  

Dynamic Matrix Predictive Control Based on RBF Neural Networks Predictive Model

作　　者：李志武[1] 

机构地区：[1]辽宁石油化工大学职业技术学院,抚顺113001

出　　处：《科学技术与工程》2010年第7期1643-1647,1651,共6页Science Technology and Engineering

摘　　要：研究了一种利用RBF神经网络预测模型的动态矩阵控制算法,首先利用动态节点生成构造性RBF神经网络辨识对象模型,同时预测对象的未来输出,然后用传统的动态矩阵控制算法进行滚动优化和反馈校正。仿真表明该算法在非线性对象的任意工作点都可以通过神经网络辨识获得工作点附近的近似线性模型,具有较好的实时性。A new Dynamic Matrix Control （DMC） algorithm using Radial Basis Function （RBF） neural networks as predictive model is proposed. First the dynamic node creating training RBF neural network is adopted for identifying model of the controlled plant and produce predictive output. Then the DMC algorithm is used in receding horizon and revising feedback. After the proposed algorithm is applied to simulate a class of nonlinear plants , the result demonstrates that the approximate linear model near work spot can be achieved in any work spots of nonlinear plant by neural networks identification and it has better real time.

关 键 词：径向基神经网络 动态矩阵 动态结点生成训练算法 非线性模型预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]